From superpixel to human shape modelling for carried object detection

作者：

Highlights：

• Carried objects are detected without assumption on their locations and their shapes.

• Carried objects can be detected from a single frame without temporal information.

• Analysing local shape information of contours can benefit from superpixels.

• Superpixel based feature extraction leads to improvement in human shape modeling.

摘要

•Carried objects are detected without assumption on their locations and their shapes.•Carried objects can be detected from a single frame without temporal information.•Analysing local shape information of contours can benefit from superpixels.•Superpixel based feature extraction leads to improvement in human shape modeling.

论文关键词：Superpixel,Shape context,Codebook,Carried object

论文评审过程：Received 31 October 2017, Revised 3 October 2018, Accepted 14 December 2018, Available online 22 December 2018, Version of Record 14 January 2019.

论文官网地址：https://doi.org/10.1016/j.patcog.2018.12.009